sign language parsing

[Antonio Carlos da Rocha Costa]

We are not yet able to parse sign language texts,
but we are already able to search signs in sign language
texts.

I think everybody would agree that searching texts is
the most basic text processing task. That's why we
started with it.

We are usign the SignWriting system

     http://www.signwriting.org

to record sign language in written form. Such documents
are stored in the SWML format

     http://swml.ucpel.tche.br

which is an XML-based format for storing SignWriting texts.

  The search procedure is very basic:

  1) it can either match signs that have the symbols in the
same absolute position in their respective sign boxes
  2) or else it can match signs considering the relative
positions of the symbols in the sign boxes, taking one of
the symbols in the signs as a reference (usually the head).

  The algorithm is very simple and not optimized. It is
implemented just to demonstrate the idea that searching
sign language texts is possible (which is obvious, as soon
as you understand the principles of the SignWriting system).

  We are a Computer Science group, not a Linguistics group.
When we enter the parsing and semantic analysis tasks (which
we will have to, sooner or later) we will need the participation
of linguists.

  Sorry for sending this first message without having introduced
myself before. I've just been reading messages in the list, for a long
time now. This is the first time I feel I can contribute something
possibly interesting.

  All the best,

  Ant ­? io Carlos


---------------------------------
Ant ­? io Carlos da Rocha Costa
Escola de Inform H² ica
Universidade Cat ¡_ ica de Pelotas
Pelotas, RS, Brazil.





> Quoting Dan Parvaz <dparvaz at UNM.EDU>:
>
>
> > Are you referring to simply parsing a signed language, or are you
> > talking about Machine Translation (in which syntactic parsing is
> > often
> > a step)?
> Yes,thank you Dan! I mean Machine Translation.
> And what do you mean by "written text?" What is the
> > original form which would be "parsed?" HamNoSys? Papaspyrou's
> orthography? A
> > video signal? Glosses? And what kinds of input are required for the
> > signing avatar?
>
> The input will be written typed text and the output will be signed
> animated video on screen. The project is a joint effort between 3D
> animation video experts and linguists. At this first point we aim to
> parse very simple sentences in GSL, still retaining as many as possible
> of the phonological characteristics of GSL,not only manual signs.
>
> How much of the phonological/phonotactics
> > (kinematics,
> > etc.) need to be pre-processed?
> >
> > I have other questions, but this is a good kick-off point. I suspect
> > there are a few of us here who can handle one or more of those
> > pieces
> >
> I have contacted one person who has had some practical experience on a
> similar project so far, and I would really appreciate any additional
> information on specific difficulties and pitfalls of such an attempt.
> The rest of the team in our project have a strong background in their
> own areas but, as the only sign linguist in the project, I suspect the
> goals might be too ambitious...
>
> Thanks for all comments,
> Galini
> > Cheers,
> >
> > Dan.
> >
> >
>




-------------------------------------------------
This mail sent through IMP: http://horde.org/imp/